Analytical Approximations to Predict Performance Measures of an Assembly Line

摘要

The nature of the web server assembly process is complex. Servers are highly customizable and hence typically are made-to-order. When a failure is observed during the testing phase, the servers have to be disassembled, reassembled, and retested; thus, job/server circulation is typical. The entire assembly process is modeled as a network of queues. Fork and join systems are used to model parallel processing. An open queuing network is considered with multiple product classes, job circulation due to failures, and resource sharing. Both the arrival and service rates are from a Markovian distribution. A parametric decomposition approach is adopted to develop analytical approximations to measure the performance of the assembly system. The analytical models are verified by comparing its results with simulation models.